Very short summary of this paper:

Very excited that my paper with @katakeith.bsky.social is now out in @polanalysis.bsky.social. We investigate whether LLMs actually follow the instructions/definitions provided in codebooks, propose some diagnostics, and release a new evaluation dataset.
www.cambridge.org/core/journal...

Currently in FirstView: “Synthetically generated text for supervised text analysis.” @ahalterman.bsky.social proposes using LLMs to generate synthetic training data for training smaller, traditional supervised text models.

I promise that training models on synthetic text is a better idea than it sounds. For the theoretically squeamish: think of this as model distillation (LLM --> small classifier). For the hardcore empiricists, there are a few F1-go-up plots.

New paper in Political Analysis on synthetic text data for training classifiers. Main idea: generate training examples with LLMs, then fit classifiers on synthetic (+real) text. Paper has validations and guidance.
Blog: andrewhalterman.com/post/synthet...
Paper: www.cambridge.org/core/journal...

I couldn't find a tutorial I liked to get students who know R up and running with Python, so I wrote my own! Part 1 is here: andrewhalterman.com/post/python_...
(And if you have a tutorial you like, please let me know!)

The optimal amount of dad joke cringe when teaching undergrads is > 0.